Do Babies Matter?: Gender and Family in the Ivory Tower

Appendix

115 This book is based on primarily two data sources. The first is the nationally representative Survey of Doctorate Recipients. The second is a series of surveys conducted at the University of California. The Survey of Doctorate Recipients For more than fifty years all new Ph.D. recipients in the United States have been administered questionnaires, composing the Survey of Earned Doctorates.1 Since 1973, slightly under 10 percent of Survey of Earned Doctorates respondents have been selected for ongoing biennial interviews that continue until age seventy-six or relocation outside the United States. Together the repeated interviews of new and former Ph.D. recipients compose the Survey of Doctorate Recipients (SDR).2 The result is a large and continually replenished set of panel data on academic careers. Overall response rates are good; for instance, 87 percent of respondents completed the survey in 1991.3 Moreover, we employ survey weights that adjust for response bias. To account for the effects of the weights and sample design on t-ratios, we report the results of significance tests based on robust standard errors.4 It should be observed that the SDR is much broader than other data sets commonly used to study higher education faculty. Traditionally the emphasis has been on bench scientists.5 Similarly, studies most often focused on academics at major research universities. The SDR, as a large representative sample of Ph.D. recipients, allows us to overcome both these limitations. In addition, as the SDR is a sample of Ph.D. recipients rather than of faculty members , we know about Ph.D.s who do not become tenure-track professors. These APPENDIX Data and Analysis 116 APPENDIX individuals provide valuable comparison groups for the analyses reported in chapters 2 and 4. Prior to 1981 the SDR did not collect information about respondents’ children , so these earlier years are not used.6 Accordingly, most of our analyses are based on SDR data starting in that year. Select analyses begin only with the 1983 data because of dependent variable coding issues. Our time series end in one of two years depending on the composition of the sample. The SDR included humanities doctorates only through the 1995 survey; thereafter they were the victims of budget cuts. Accordingly, we present results that extend through 1995 based on all academic disciplines, and results extending through 2003 based on only the sciences (including the social sciences). One great strength of the SDR is its large sample sizes. By the 2003 survey cycle more than 160,000 doctorate recipients had participated.7 The smallest SDR sample used in any of our multivariate analyses is over 5,500; the largest, over 30,000. Sample sizes for all SDR analyses are shown in table A.1. Missing data are deleted listwise, except when large numbers of missing cases (that is, sufficient to allow estimation with dummy variables for missing data) may represent substantively meaningful differences between respondents. For these items, including race, time to complete Ph.D., quality of degree-granting institution, and type of employing institution (all categorical variables), we code additional dummy variables for missing data and, for continuous variables, mean imputation with a missing-data dummy. More sophisticated means of handling missing data, such as multiple imputation , do not produce appreciably better estimates of regression coefficients and standard errors.8 Finally, all reported SDR results pertaining to gender, marriage, divorce, and children are statistically significant unless otherwise noted. University of California Surveys We use data from three separate surveys administered at nine schools of the University of California system (the tenth, UC Merced, opened in 2005 and did not participate); a fourth survey was limited to UC Berkeley. The three UC-wide surveys were administered under the auspices of Mary Ann Mason and were fielded under the imprimatur of the president of the UC system; Marc Goulden, Angelica Stacy, and Sheldon Zedeck conducted the fourth survey. Marc Goulden and others designed the surveys, which were intended to explore work-family issues at UC. The surveys, which were administered via the Internet, produced both qualitative and quantitative data. The first survey targeted UC tenure-stream faculty and was fielded in 2002 and 2003.9 It had a response rate of 51 percent and produced 4,460 responses. DATA AND ANALYSIS 117 The second survey was administered to UC doctoral students in 2006 and 2007.10 It had a response rate...